|Home | About | Journals | Submit | Contact Us | Français|
In 1983, reports of antibodies in subjects with major depressive disorder to an as-yet uncharacterized infectious agent associated with meningoencephalitis in horses and sheep led to the molecular cloning of the genome of a novel, negative-stranded neurotropic virus, Borna disease virus (BDV).1,2 This advance enabled the development of new diagnostic assays including in situ hybridization, PCR and serology based on recombinant proteins. Since these assays were first implemented in 1990 more than 80 studies have reported an association between BDV and a wide range of human illnesses that include major depressive disorder, bipolar disorder, schizophrenia, anxiety disorder, chronic fatigue syndrome, multiple sclerosis, amyotrophic lateral sclerosis, dementia and glioblastoma multiforme.3,4 However, to date there has been no blinded case-control study of the epidemiology of BDV infection. Here, in a United States-based, multi-center, yoked case-control study with standardized methods for clinical assessment and blinded serologic and molecular analysis, we report the absence of association of psychiatric illness with antibodies to BDV or with BDV nucleic acids in serially-collected serum and white blood cell samples from 396 subjects, a study population comprised of 198 matched pairs of patients and healthy controls (52 schizophrenia/control pairs, 66 bipolar disorder/control pairs, and 80 major depressive disorder/control pairs). Our results argue strongly against a role for BDV in the pathogenesis of these psychiatric disorders.
The concept that infectious agents may cause mental illness has a long and complex history.5,6 Although acute brain infections (e.g., with herpes simplex or rabies virus) are clearly associated with sporadic cases of psychosis,7–9 and others like influenza are historically associated with epidemic psychosis,10–12 no linkage is established between infection and the majority of cases of affective disorders or schizophrenia. Infection-based models have often focused on specific microbes as potential culprits. One candidate is Borna disease virus (BDV), a noncytolytic RNA virus that causes movement and behavior disturbances in warmblooded animal hosts from birds to primates.13–15 BDV has been the subject of intensive research interest over the past two decades,3,4 but findings are inconsistent. Recent reports of novel bornaviruses in birds (avian bornavirus [ABV])16,17 and discovery of BDV sequences in genomes of mammals, including humans,18,19 have accentuated this interest.
Some studies note an increased prevalence of antibodies to BDV proteins in psychiatric patients, but rates vary substantially by patient groups (1.6–100%) and assay type. High seroprevalence rates are also found in controls (1.2–46%).3,13 Serial analyses may yield higher seroprevalence rates,20–22 though some studies fail to find differences between serially-assessed patients and controls.21 Some variability may be attributed to assay specificity. In chronic fatigue syndrome, 27 of 169 subjects with disease were seropositive by enzyme-linked immunosorbent assay (ELISA), but none (0/53) were positive by western immunoblot (WIB).23 The highest reported rates in psychiatric patients were found using methods designed to detect BDV-specific antigen-antibody complexes (up to 100% in affective disorder samples);24 however, other investigators report these methods are nonspecific.25 Results of studies using molecular strategies are also inconclusive; however, differences in results are more difficult to ascribe to assay specificity given that research groups tend to use similar primer sequences and protocols.
Naturally-infected horses, sheep, cattle, cats and birds could serve as reservoirs for the virus; however, there are no detailed epidemiologic studies in animal populations and no studies demonstrating transmission from domestic animals to humans. BDV is transmitted efficiently through contact with nerve terminals (e.g., olfactory infection), but the observation that BDV may be present in peripheral blood cells suggests the potential for hematogenous infection. One study in Japan revealed that 4–5% of random blood donors had BDV nucleic acids in peripheral blood mononuclear cells.26 The only study to report a higher prevalence of historical infection in healthy blood donors (30%) relied on a sandwich enzyme immunoassay to detect circulating immune complexes (CIC-EIA24) that does not discriminate the presence of BDV antigen from nonspecific reactivity.27 Limitations in our understanding of pathogenesis continue to restrict our ability to tailor study design for selection of best sampling compartment, timing of sample collection relative to illness onset or exacerbation, and diagnostic markers for detection of infection.3,28
This multi-center, case-control investigation addressed these research gaps by subjecting prospectively-collected samples from well-characterized patients with neuropsychiatric diseases and matched, healthy controls to blinded analysis using standardized methods for serologic and molecular analysis. Our primary goal was to test for an association between infection and neuropsychiatric disorder, as determined by antibody and/or nucleic acid status. We also evaluated whether infection is associated with specific clinical variables, using rigorous methods for diagnosis and assessment. The prevalence of serum antibodies to BDV and of BDV nucleic acids in peripheral white blood cells (WBC) was determined in patient-control pairs, matched on age, sex, geographic residence, socioeconomic status, and timing of blood draws.
Patients with DSM-IV diagnoses29 of schizophrenia (SZ), bipolar disorder (BD) and major depressive disorder (MDD), ages 20–75 years, were recruited through outpatient clinics of University of California Los Angeles (UCLA) and University of California Irvine (UCI) using informed consent procedures and protocols approved by the Institutional Review Boards of UCLA, UCI, and Columbia University. We hypothesized that if a virus were associated with psychiatric disorder, evidence of infection would be detectable in peripheral blood during acute onset or exacerbation of pre-existing psychiatric illness, with peak virus-specific, IgG antibodies during convalescence, 6 weeks later. In keeping with this paradigm, for study inclusion, patients had to experience psychiatric illness onset or exacerbation within 6 weeks of study entry (T1) and be available for repeat evaluation and blood draw 6 weeks later (T2). Patients with unstable medical illness were excluded. Recruited patients nominated healthy social contacts, to serve as matched controls, of the same sex, same socioeconomic status, similar age (± 5 years), and residing in the same geographic region as the patient. Yoked control candidates who were blood relatives or current or previous household or sexual partners of nominating case subjects, or who had unstable medical illness, were excluded. Controls with substance use disorders were excluded, but substance use disorders were not an exclusion for patients except for those with histories of intravenous drug abuse.
Diagnoses of SZ, BD, and MDD were established by Structured Clinical Interview for DSM-IV Axis I Disorders (SCID).30 Illness severity ratings were derived using Clinical Global Impressions-Severity31 and Global Assessment of Function scales.29 The SCID was also used to establish presence of substance use disorders in patients and any exclusionary Axis I and Axis II diagnoses. Controls were screened using the Depression Rating Scale32,33 for MDD and BD subjects; severity of mania was evaluated through the Young Mania Rating Scale.34 SZ severity was rated using the Positive and Negative Symptoms Scale.35 Semi-structured interviews provided standardized collection of exposure and demographic data; individual and family medical and psychiatric histories; medication history, including response to prior neuropharmacologic trials;36–38 and current medications. At T2, severity ratings were repeated, and changes from T1 in physical status (intercurrent illness) or medications were recorded.
Blood samples were first collected from patients within 6 weeks of onset of an acute episode or clinically significant exacerbation (T1), and 6 weeks later (T2) to allow for determination of changes in viral nucleic acid load or antibody titers. T1 and T2 blood samples were collected from controls within 4 weeks of the respective T1 and T2 blood draws from the case to whom that control was yoked. Blood was collected into red-top tubes to obtain serum, and into blue-top tubes (3.2% sodium citrate) to obtain plasma and WBC. All samples were coded to protect patient privacy and ensure blinded analysis. Samples were processed within 24 hours of collection in a facility never used for BDV-related work (Supplementary Information).
To minimize potential for artifacts due to inadvertent introduction of BDV nucleic acid, PCR master mixes were prepared in different rooms than those employed for PCR, and separate glove boxes and equipment were used for individual work steps (master mix, template addition, handling of positive/negative controls). Quantitative real-time reverse transcriptase (RT)-PCR assays for BDV nucleoprotein (N) and phosphoprotein (P) gene sequences: Real-time RT-PCR assays were established for detection of strain V, He/80, and the newer isolate, No/98, which represent the most divergent BDV sequences. PCR primers for BDV N and P gene sequences of the designated strains, and for porphobilinogen deaminase (housekeeping gene), were designed using Primer Express 1.0 software (Applied Biosystems, Foster City, CA). Template standards were cloned by PCR into vector pGEM-T easy (Promega, Madison, WI) using viral RNA obtained from cultured cells infected with strain V, He/80 or No/98.
To allow recognition of amplification products from synthetic template standards used in real-time RTPCR assays, selective mutations were introduced to create specific restriction sites. See design of plasmids and restriction sites for use with He/80 and strain V primers and probes, and No/98 primers and probes, in Supplementary Information
RNA was extracted from WBC using Tri-Reagent (MRC). RNA quantity and quality were assessed by spectrometry and optical density (OD) 260/280 ratios. cDNA was synthesized from 2 μg of RNA from each sample using Taqman reverse transcription reagents (Applied Biosystems) (66 μl/well). Five μl of RT reaction volume (cDNA representing 150 ng of starting RNA) was combined with 2.5 μl of each primer and probe, and 12.5 ul of TaqMan Universal Master Mix (Applied Biosystems) and subjected to the following thermal cycling conditions (Model 7700 Sequence Detector, Applied Biosystems): 2 min, 50°C; 10 min, 95°C followed by 45 cycles of 1 min, 60°C and 15 sec, 95°C. Each PCR plate included dilution rows of the respective cloned BDV plasmid standard as well as the cloned housekeeping gene standard, and negative controls representing normal human leukocyte RNA (added at RT stage; BD Biosciences Clontech, San Diego, CA) and human placenta DNA (added at PCR stage; Sigma, St. Louis, of input RNA and to normalize gene expression values for target genes.
RNA from coded T1 WBC aliquots was tested in duplicate for presence of P and N gene transcripts in randomized, blinded fashion by two separate real-time RT-PCR analyses, using 2 different primer sets. Prior to beginning PCR assays, criteria for establishing results as positive were defined as consistent detection of either P or N gene transcripts, using at least one of the 2 primer sets. In the event that discordant results were observed for the first two aliquots assessed by real-time RT-PCR, RNA extracted from a third WBC aliquot would be subjected to an additional round of real-time RT-PCR analysis. In this instance, samples would be designated positive or negative based on the two real-time RT-PCR results found to be concordant (e.g., 2 of 3 negative real-time RT-PCRs would be designated a negative result; 2 of 3 positive real-time RT-PCRs would be designated a positive result).
Consensus PCR for ABV sequences: Following identification of multiple novel bornaviruses in diseased psittacines,16,17 we established new primer sets and assays for detection of ABV genotypes 1–4 (Supplementary Information).
PCR products were size-fractioned on agarose gels, purified (Qiaquick PCR purification kit; QIAGEN, Valencia, CA) and directly dideoxy-sequenced (Genewiz, South Plainfield, NJ). Sequences were analyzed by GenBank BLASTn nucleotide search, comparison to a 440-nt sequence for ABV 1–4 (GenBank accession numbers FJ169441, FJ169440, EU781967, GU249595), and alignment using Molecular Evolutionary Genetics Analysis (MEGA) software version 22.214.171.124
ELISA: Coded serum samples were assessed for presence of antibodies to BDV proteins by ELISA as previously described.40 Recombinant BDV proteins p40 (N) and p23 (P) were expressed from pET-BDV N and pET-BDV P plasmids (vector pET15b, Novagen, Gibbstown, NJ) as His-tagged fusion proteins for use as antigens in ELISA serology experiments. Carbonic anhydrase (Sigma, St. Louis, MO) served as negative control antigen. All proteins were confirmed as immunoreactive in ELISA with sera from experimentally-infected rodents (rats) and primates (Rhesus macaques) before use in assays with human materials. Human T1 and T2 samples were tested in duplicate in 96-well microtiter plates (dilutions of 1:1000, 1:2000, 1:4000, 1:8000) using 80 ng/well of recombinant N, P, or carbonic anhydrase (nonspecific protein target). Sera from experimentally-infected Rhesus macaques were used as positive control (1:8000 dilution) and normal human serum (Jackson Laboratories, West Grove, PA) as negative control (1:1000 dilution). Negative control wells without primary antisera were included for calibration. Serum samples were considered reactive with a BDV protein if OD reading at 450 nm was two standard deviations above the mean of their nonspecific reactivity with carbonic anhydrase. Sera positive by initial ELISA for either or both BDV proteins were subjected to a second round of ELISA. Sera positive in the first round of ELISA but not reactive with the same proteins in repeat ELISA were considered negative. Samples were only considered positive in ELISA if immunoreactive to both BDV N and P proteins; any sera found positive by ELISA would be subjected to WIB analysis. Reactivity with both BDV N and P proteins by WIB was required to confirm infection. Any sera positive by initial ELISA but not reactive with the same proteins in WIB would be considered negative. If positive sera were found, these were serially-diluted and retested in ELISA to establish antibody titers against individual BDV proteins (ELISA titer defined as endpoint dilution yielding OD of 0.3).
Immunofluorescence assay (IFA): Baseline serum samples were analyzed by IFA in a multi-step strategy. Each test series was conducted in 24-well plates and included two human control serum samples defined in previously-published work as positive and negative for anti-BDV antibodies.41 Two independent raters evaluated immunoreactivity of coded samples at each step. Samples were initially evaluated in a screening IFA (1:10 dilution) using C6 cells infected with BDV strain He/80. Subsequently, in a second round of screening, samples rated positive in the first step were tested (1:10 dilution) with both BDV-infected and uninfected C6 cells. Immunoreactivity of samples with infected C6 cells was compared with immunoreactivity with uninfected C6 cells to determine specificity of binding to BDV proteins vs. nonspecific nuclear staining; samples with punctate immunoreactivity observed only in infected C6 cells, and not in uninfected C6 control cells, were considered confirmed as screen-positive samples. All samples rated as positive in these two IFA screening steps were retested using infected and uninfected C6 cells to determine titer and avidity (immunoreactivity with 6 M urea vs. no urea). To assess consistency of results across multiple IFA assays, a set of 43 samples, comprising additional aliquots of samples characterized as positive as well as negative in the first two screening steps, was subjected to repeat IFA testing.
PASW Statistics 18.0.3 (IBM SPSS Statistics/Mac, Armonk, NY) was used for all statistical analyses. Distributions were examined to ensure they did not deviate from normality; data meeting this criterion were used to derive paired-difference scores and evaluated for each variable across diagnostic groups using one-way ANOVA. Diagnosis-restricted patient/matched control group status served as the independent variable. For continuous data deviating from normal distributions, group comparisons and correlational analyses were conducted using Mann-Whitney U (nominal α=0.05). Chi-square analyses were performed for nominal data (two-tailed Fisher’s Exact Test for significance).
The final study sample consisted of 198 matched case-control pairs (396 subjects) with data available at T1: 52 SZ and 52 matched controls (SZ-C); 66 BD, and 66 BD-C; and 80 MDD and 80 MDD-C (Table 1). One-hundred fifty-one pairs (302 subjects) also had T2 data available (48 SZ/SZ-C, 51 BD/BD-C, and 52 MDD/MDD-C matched pairs). Mean age of subjects did not differ between or across diagnostic groups. The proportion of males was higher in SZ (82.7%) (and hence their matched controls, 82.7%) than in the other patient groups (BD, 56.1%; MDD, 57.5%) or their respective matched control groups (BD-C, 56.1%; MDD-C, 57.5%; p=0.004). Distributions of ethnicity, country of origin and measures of socioeconomic status (% below poverty level and mean household income based on residence zip code) were similar across groups (Table 1). Patients did not differ from respective matched controls in exposure to domestic pets; current use of caffeine, tobacco, and recreational drugs; and foreign travel (Table 1). SZ subjects were less likely to consume alcohol than SZ-C subjects (21.2% vs. 42.3%; p=0.037).
RNA from all T1 WBC samples were studied in multiple real-time RT-PCR assays designed for detection of He/80, strain V and No/98 strains of BDV with sensitivity thresholds of less than 25 RNA copies (Supplementary Information). None of the samples showed evidence of N or P gene sequences by any of the real-time RT-PCR primer sets employed (Table 2).
Using a newly-designed PCR assay for detection of M and P gene sequences of ABV genotypes 1–4 with a sensitivity threshold of less than 100 RNA copies (Methods, Supplementary Information), we examined WBC-derived RNA samples from a subset of 97 patients (24 SZ, 33 BD, 40 MDD) with current (last 6 months) or historical exposure to birds and/or fowl (46 of 97, or 47.4%, or 51 of 97, or 52.5%, respectively). No sequences specific for ABVs were identified (Table 2).
All T1 and T2 serum samples were tested in duplicate for immunoreactivity to recombinant N, P, and carbonic anhydrase (nonspecific protein target). No samples were positive for antibodies to BDV by ELISA at either sampling time point.
T1 serum samples were tested for binding to He/80-infected C6 cells. Eight sera were consistently immunoreactive (8 of 396, or 2%). These sera were tested for avidity by repeating the assay in the presence of 6 M urea. Four sera had high avidity (2 BD, 1 BD-C, 1 SZ-C). Four had low avidity (1 BD, 2 BD-C, 1 MDD).
Analysis of 198 patients with neuropsychiatric disease and their 198 pairwise-matched controls revealed no evidence of BDV association with psychiatric diagnosis at acute onset or exacerbation of illness, using a broad repertoire of molecular and serologic assays. Additionally, we found no molecular evidence of BDV infection in any subject. Serology varied by platform. ELISA was negative for all subjects at T1 and T2. IFA indicated the presence of high avidity antibodies in 4 subjects: 2 BD, 1 BD-C, and 1 SZ-C. However, follow-up WIB analysis found no immunoreactivity of any of these four samples to BDV N or P (data not shown).
Disparities between our results and prior studies may reflect differences in subject recruitment and sampling procedures. First, patients and controls were of similar age (+ 5 years), sex, ethnic/racial group, socioeconomic status, geographic residence, and date of blood sampling, controlling for common factors potentially confounding both host risk of exposure as well as response of each individual to infection. Second, participants were rigorously diagnosed to validate presence or absence of psychiatric illness (presence of the relevant DSM-IV diagnosis in SZ, BD and MDD groups, without other primary psychiatric disorders; absence of psychiatric and substance use disorder in controls) and provide information about phenotypic differences that might be linked to altered viral risk (seasonality, melancholia, psychosis, treatment-resistance, course of illness). Potential exposures (e.g., animals, substances, foreign travel) and risk factors for altered response to infection (personal or family medical history, including autoimmune disorders; medications) were also thoroughly-characterized with standardized instruments. Third, the serial sampling incorporated into our study design - within 6 weeks of an acute flare of illness and again 6 weeks later - were uniquely informed by classical acute and chronic phase models of host response to viral infection. The study was thus poised to collect critical information that might shed light on the influences of cofactors on BDV measures, including time and clinical severity. It is also possible that North American subjects differ from other study populations in prevalence of exposure to bornaviruses for reasons as yet unknown.
An alternative explanation for differences in results obtained in this and other studies may be our methods for sample collection, processing, and assay protocols designed to minimize risk of sample contamination. Given the high degree of sequence conservation across BDV strains, and the resultant difficulties in distinguishing bona fide BDV sequences, all blood samples were processed in an independent laboratory using newly-purchased tools and equipment devoted to this project, and stored in a dedicated freezer. The incorporation of more than one strategy for serologic detection of antibodies to bornaviral proteins (ELISA, IFA), and inclusion of methods to measure avidity of detected antibodies (urea treatment in IFA), minimized the possibility that sensitivity of any particular assay was insufficient for detection of low levels of anti-BDV antibodies. Establishment of assay-specific study definitions of positive and negative findings, prior to conducting any assays, mitigated against potential concerns regarding bias in assay interpretation.
All sera were negative by ELISA; the majority were negative by IFA. The four serum samples with positive signal by IFA were shown to be nonspecific in follow-up WIB studies. Immune complex analyses24 performed in another laboratory showed no association between the presence of complexes and disease status (data not shown). Real-time RT-PCR of WBC derived RNA, including samples from subjects with IFA signal, showed no evidence of BDV-related nucleic acids or ABV sequences. These findings are consistent with those of Wolff and colleagues wherein neither BDV antigen nor nucleic acids were detected in human blood samples previously reported as positive in circulating immune complexes enzyme-linked immunoassay.25 The greater consistency in the current study across multiple assay platforms and detection targets may relate to its large size, well-controlled study design, use of serial samples, efforts to minimize risk of laboratory contamination, and sample blinding strategy. We cannot exclude the possibility that BDV RNA or antibodies may be present in the cerebrospinal fluid (CSF) of some patients with neuropsychiatric diseases. However, there are no replicated studies that show evidence of infection in brain or CSF. The vast majority of reports of footprints of BDV infection in these diseases are based on analyses of PBMC and sera. Recent reports identifying the incorporation of BDV sequences into the genome of humans and other mammals18,19 indicate that bornaviruses infected primates over 40 million years ago. Whether BDV has ever been pathogenic for humans, or altered function in clinically significant ways, remains unsolved. Nonetheless, using sensitive laboratory measures, this large, tightly-controlled, case-control analysis provides powerful evidence that molecular and serologic markers of BDV exposure are largely absent from the peripheral blood of all subjects, even during acute neuropsychiatric illness, and argues strongly against a role for BDV in the pathogenesis of schizophrenia or mood disorders.
Supplementary information is available at Molecular Psychiatry’s website.
Work reported here was supported by NIH award MH57467 (WIL). We thank Meera Bhat for data derived from linkage to the United States Census Bureau database, and Vishal Kapoor for assistance with Western immunoblot analyses.